Re: Proposed API extension for Fido U2F devices

On Sun, Feb 9, 2014 at 12:11 PM, Siva Narendra <siva@tyfone.com> wrote:

> Hi Juan and everyone.
>
> Smart Card support which is already a hardware standard and it is well
> know what language it speaks (ISO7816) should be considered as an
> alternative to new and yet to be common U2F hardware standard.
>
U2F was designed with the web and the web security requirements in mind.

I do not believe it would at all be appropriate for security to consider
exposing ISO7816-devices to the web.


>  As you all know smart cards are not only already a standard, it has sold
> in the billions and costs as little as $0.60.
>
> In addition,  Public key methods (without using PKI) is already well
> supported with repurposed smart card security applets such as PKCS #15 that
> can work with existing server infrastructure with very minimal change, to
> enable for eg mutually authenticated TLS connection with no need for
> 3-party PKI. We already have this working in Firefox where the smart card
> is connected to the device in one of many possible ways.
>
Very few smart cards support PKCS#15. It's well understood that a variety
of card edges exist - often custom and proprietary.

Further, as has been discussed repeatedly in the past, the security model
of TLS client certificate authentication - where neither party directly
controls the resulting message being signed and reliance on a 'trusted'
(read: installed) program (the UA and/or middleware, depending on OS API
used) is used to perform the signing - is _vastly_ different than allowing
arbitrary JS access to a smart card.


>  Not to mention smart card applets are extensible to any future standard
> such as the yet to be released NIST standards on derived credentials for
> government use.
>
> In addition smart cards are the only security hardware that have well
> defined and well followed security certification across multiple industry
> verticals.
>
> It would be unproductive to consider hardware without smart cards as part
> of it.
>
I strongly believe that it would be far more unproductive to consider smart
cards, in their vast, non-standard specific use cases and distinct security
models and risks, versus something with a very defined, limited scope, and
with strong browser/industry involvement on defining something actually
suitable for the web.

>  Dear Harry - What is the expected timing of considering hardware
> extension? We will take ownership in writing an alternative to U2F that is
> smart card based.
>
If and when this WG is rechartered, we can revisit whether or not hardware
tokens are in scope. The biggest determinant for deciding whether or not
they are in scope will require a demonstration that the many security
issues that have been highlighted during the F2F and on the list can be
sufficiently resolved. This seems extremely unlikely for the general smart
card case, whereas the U2F proposal clearly demonstrates attention to these
concerns (even though it can use refinement/improvement)

Cheers,
Ryan

>  Best regards,
> Siva
>
>
> *--*
>
>
> *Siva G. Narendra Ph.D.  CEO - Tyfone, Inc.Portland | Bangalore | Taipei
> www.tyfone.com <http://www.tyfone.com>*
> *Voice: +1.661.412.2233 <%2B1.661.412.2233>*
>
>
>
> On Sun, Feb 9, 2014 at 2:04 AM, Harry Halpin <hhalpin@w3.org> wrote:
>
>>  On 02/04/2014 10:41 PM, Juan Lang wrote:
>>
>> Hi folks,
>> I'm aware that hardware-backed keys are out of scope for the current
>> round of WebCrypto work, so I don't expect this to be ready for
>> standardization for some time. Nevertheless, I've got a proposed extension
>> to WebCrypto to support Fido Alliance (fidoalliance.org) universal
>> second factor (U2F) devices:
>>
>> https://docs.google.com/a/chromium.org/document/d/1EEFAMIYNqZ7XHCntghD9meJwKgNOX7ZN-jl5LJQxOVQ/edit#
>>
>>  I apologize that the proposal may lack some context, like, just what is
>> a U2F device, and what language does it speak? I promise update it with
>> pointers to public docs once they are made public. In the meantime, I'll
>> act as a poor substitute by answering questions myself, either in the doc
>> or in email.
>>
>>  I'd appreciate any feedback you might have. Thanks very much,
>>  --Juan
>>
>>
>> I haven't had to look at this in detail, but upon first look it seems
>> sensible. The general direction is one that the W3C is actively interested
>> in. While this would be outside the current charter, we will re-charter the
>> Working Group once the current version of WebCrypto (at earliest) has
>> exited Last Call and working with FIDO Alliance would likely be mutually
>> beneficial.
>>
>>     cheers,
>>       harry
>>
>>
>